Spatio-temporal evolution of wet-dry event features and their transition across the Upper Jhelum Basin (UJB) in South Asia
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022
https://doi.org/10.5194/nhess-22-287-2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.
Spatio-temporal evolution of wet–dry event features and their
transition across the Upper Jhelum Basin (UJB) in South Asia
Rubina Ansari and Giovanna Grossi
Department of Civil, Environmental, Architectural Engineering and Mathematics, University of Brescia, Brescia, Italy
Correspondence: Rubina Ansari (r.ansari@unibs.it)
Received: 13 September 2021 – Discussion started: 17 September 2021
Revised: 5 December 2021 – Accepted: 23 December 2021 – Published: 3 February 2022
Abstract. The increasing rate of occurrence of extreme 1 Introduction
events (droughts and floods) and their rapid transition mag-
nify the associated socio-economic impacts with respect There is growing evidence that recent warming is leading to
to those caused by the individual event. Understanding of significant alteration in the hydrological cycle, exacerbating
spatio-temporal evolution of wet–dry events collectively, extreme weather events in general (Peterson et al., 2012) in
their characteristics, and the transition (wet to dry and dry many regions of the world. Extreme weather events such as
to wet) is therefore significant to identify and locate most floods and droughts and their rapid successions (recurrent
vulnerable hotspots, providing the basis for the adaptation spells) during the past few decades have taken a heavy toll on
and mitigation measures. The Upper Jhelum Basin (UJB) both life and property. Moreover, such events can have large
in South Asia was selected as a case study, where the rel- impacts on water availability, agriculture and food security,
evance of wet–dry events and their transition has not been power production, and natural ecosystems (He et al., 2019;
assessed yet, despite clear evidence of climate change in the Sheffield and Wood, 2012). These events are projected to re-
region. The standardized precipitation evapotranspiration in- gionally intensify and be more frequent within the context
dex (SPEI) at the monthly timescale was applied to detect of global warming, underscoring the importance of research
and characterize wet and dry events for the period 1981– on wet–dry extreme weather events collectively. The climate
2014. The results of temporal variations in SPEI showed change projections for the Asian continent in the sixth As-
a strong change in basin climatic features associated with sessment Report (AR6) of the Intergovernmental Panel on
El Niño–Southern Oscillation (ENSO) at the end of 1997, Climate Change (IPCC) reported that during the 21st cen-
with the prevalence of wet and dry events before and af- tury South Asia is likely to face more intense and frequent
ter 1997 respectively. The results of spatial analysis show a heatwaves and humid heat stress, whereas both annual and
higher susceptibility of the monsoon-dominated region to- summer monsoon precipitation will increase, with enhanced
wards wet events, with more intense events occurring in the inter-annual variability (medium confidence) (Arias et al.,
eastern part, whereas a higher severity and duration are fea- 2022). Various studies at local, basin, national, and regional
tured in the southwestern part of the basin. In contrast, the scales already documented and acknowledged the vulnerabil-
westerlies-dominated region was found to be the hotspot of ity to climate change of that region (He and Sheffield, 2020;
dry events with higher duration, severity, and intensity. More- Zhao et al., 2020; Visser-Quinn et al., 2019; He et al., 2017).
over, the surrounding region of the Himalaya divide line and Typically, wet and dry events are generally considered in-
the monsoon-dominated part of the basin were found to be dependently in water resource management and planning.
the hotspots of rapid wet–dry transition events. However, these events are inherently interconnected and gov-
erned by the same underlying hydrological processes and
atmospheric dynamics, which may augment hydro-climatic
variability under the influence of climate change (He and
Sheffield, 2020). A number of rapid wet–dry events in the
last decade acknowledged the relevance of sequences of wet
Published by Copernicus Publications on behalf of the European Geosciences Union.
288 R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features and dry events. For example, California’s large-scale flood ered in hydrology (Kourgialas, 2021). The calculation of SPI event in 2017 occurred at the offset of prolonged drought and SPEI is mathematically similar, but it differs in the input (2011–2016) (He et al., 2017; NOAA National Centers for parameters. The SPI only uses precipitation, whereas SPEI is Environmental Information, 2018). South Carolina observed based on the climatic water balance. Many studies advocate an abrupt transition (within a week) from drought to flood the use of SPEI, rather than SPI, due to its link to potential in September 2015 (He and Sheffield, 2020). Other exam- evapotranspiration (PET), which makes it more sensitive in ples include the successive drought and flood events of 2010– the context of global warming (Himayoun and Roshni, 2019; 2012 and 2015–2016 in the UK (Parry et al., 2013) and Tas- Yao et al., 2018; Huang et al., 2017; Vicente-Serrano et al., mania, Australia respectively (CSIRO, 2018). Such abrupt 2010). flood–drought transitions pose a substantial risk for water In this study, attempts were made to understand the re- management practices, especially for reservoir operation, as gional evolution of wet–dry events collectively, their charac- a trade-off should be set between short-term flood control and teristics, and their transition (wet to dry and dry to wet) for long-term water storage imperatives to satisfy water demand different severity levels ranging from moderate to extreme. (He and Sheffield, 2020). This has aroused widespread con- Here, the term “wet and dry events” does not necessarily im- cern in the scientific community to understand the wet–dry ply observed flood and drought events, unless explicitly men- interplay under a changing environment. tioned. There exists a basic difference between a flood and a During the past few decades, significant effort was put for- wet event. The former has a short duration effect (e.g. a few ward towards the adoption of a multi-hazard approach (con- hours or days) while the latter is regarded as a long period sideration of both types of extreme hydrological conditions without precipitation shortage (e.g. several months or years) at the same time) in developing resilience to climate change. (Wu and Chen, 2019). Kourgialas (2021) analysed floods and droughts collectively The proposed framework was implemented with reference in the Mediterranean agricultural region and proposed water- to the Upper Jhelum Basin (UJB), where the relevance of saving and flood protection measures for adapting to the wet–dry events and their transition have not been assessed inevitable adverse effects of climate change. Visser-Quinn yet, despite clear evidence of climate change in the region. et al. (2019) identified hotspot regions in the UK where a The UJB is located in the western Himalaya and shared spatio-temporally concurrent increase in the number of flood by Pakistan and India. The region already witnessed an in- and drought events was projected. Zhao et al. (2020) investi- crease in extreme hydro-meteorological events in the last few gated the rapid transition of flood and drought events under decades, but these events are expected to become even more present and future climate change in the Hanjiang Basin and pronounced in the coming future (Pachauri et al., 2014). A found more frequent drought-to-flood rapid-transition events study conducted over the northern highlands of Pakistan in- of higher intensity in the 21st century. Other examples in- vestigated the trends in time distribution patterns (TDPs) clude the analysis of rapid drought-to-flood transitions in and return periods for event-based extreme precipitation for river basins in China (Yan et al., 2013) and in England and the period of 1961 to 2014 and found maximum values of Wales (Parry et al., 2013). These studies employed the peak- 20- and 50-year return levels of TDP for the UJB (Zaman over-threshold (POT) method and various indices recom- et al., 2020). Another study conducted on a portion of the mended by the World Meteorological Organization (WMO) UJB located in Kashmir, India, uses the SPEI for spatio- for the detection and characterization of extreme wet and dry temporal characterization of drought events only (Himayoun events (floods and droughts). and Roshni, 2019). Akhtar et al. (2020) investigated the cor- Some commonly used indices are the standardized pre- relation of meteorological and hydrological drought using cipitation index (SPI) (McKee et al., 1993), standard- the SPEI and the standardized streamflow index (SSI) over ized precipitation evapotranspiration index (SPEI) (Vicente- the Upper Indus Basin (UIB), including UJB. They validated Serrano et al., 2010), Palmer drought severity index (PDSI) the results with a historically prolonged drought event ob- (Palmer, 1965), normalized difference vegetation index served in Pakistan (1999–2002). Another study employed the (NDVI) (Tucker, 1979), standardized drought indices (SDI) locally weighted SDI and compared it with SPI and SPEI on (Svoboda and Fuchs, 2016), and standardized anomaly index 10 meteorological stations within Pakistan (Ali et al., 2019). (SAI) (Katz and Glantz, 1986). Among these indices, SPI and Ullah et al. (2021a) evaluated four reanalysis products for SPEI are more widely accepted for the following reasons: (a) drought assessment in Pakistan using SPI and SPEI at multi- simple to calculate; (b) require few input data (precipitation ple timescales. All above-mentioned studies put a focus on and temperature), that are easily accessible in most cases; drought event characteristics only, whereas the wet events (c) standardized indices, which facilitate the comparison of and transition of wet–dry events were overlooked. This study different climatic zones; and (d) can be calculated at mul- attempts to fill this gap by addressing the following specific tiple timescales, depending on the objective. For instance, points. SPI and SPEI at short timescales (1, 2, 3, or 6 months) bet- ter reflect the meteorological and agricultural drought, while 1. How does climate change influence the evolution of the longer timescales (12, 24, or 48 months) are usually consid- regional wet–dry events? Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022 https://doi.org/10.5194/nhess-22-287-2022
R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features 289
2. How comparatively frequent were wet or dry events in Major extreme events witnessed by the basin are primar-
the past? ily led by vigorous interactions of moisture-laden monsoon
circulation and southward-penetrating mid-latitude westerly
3. What is the average transition time of wet-to-dry and troughs into the Himalayan region (Vellore et al., 2016).
dry-to-wet events?
4. Which parts of the basin are hosting hotspots for rapid
3 Data description
wet–dry transition events?
The most widely used index, SPEI, is here adopted to detect The daily observed precipitation and temperature data of 15
and characterize wet and dry events of different severity lev- climatic stations located within the political boundary of Pak-
els (moderate, severe, and extreme). The analysis was carried istan were collected from the Pakistan Meteorological De-
out both at each grid cell and averaged over the basin, using partment (PMD) and Water and Power Development Author-
corrected ERA5 precipitation and observed temperature data ity (WAPDA). For the Indian region, Indian Meteorological
for a period of 35 years (1981–2014). Department (IMD) daily gridded precipitation and tempera-
ture datasets, derived from a dense network of meteorologi-
cal stations for the Indian mainland (Pai et al., 2014), were
2 Characterization of the study area extracted at five stations and used for that region. The anal-
ysis was carried out for a period of 34 years (1981–2014),
The Upper Jhelum Basin (UJB) has a latitudinal extent due to the availability of observed data. In fact there are
stretching from 73◦ 070 to 75◦ 400 E and latitudinal extent only a few climatic stations where data are available starting
from 33◦ 000 to 35◦ 120 N (Fig. 1). The basin is mainly located from 1971, but the number of stations would not be enough
in the sub-tropics and partially in a temperate region. The for the spatial analysis. The observed temperature data were
basin drains the foothills of the western Himalaya and Pir used to calculate potential evapotranspiration (PET) using
Panjal mountains and feeds the second largest reservoir of the Thornthwaite equation (Thornthwaite, 1948) due to data
Pakistan, the “Mangla Reservoir”. The total area of the basin limitation. A study conducted by Beguería et al. (2014)
is about 33 342 km2 . The elevation ranges from nearly 223 m compared the SPEI values calculated with three different
in the southwest to about 6201 m in the north, with mean el- methods (Penman–Monteith, Hargreaves, and Thornthwaite)
evation of 2353 m a.s.l. Approximately 0.75 % (252 km2 ) of and found small differences in humid regions. Mavromatis
the basin is covered by perennial glaciers in the north of the (2007) also reported similar outcomes of PET methods for
basin (Consortium and Inventory, 2017). Grass, forest, and drought index calculation. Afterwards PET values were in-
agriculture are the three major land use–land cover (LULC) terpolated at 0.25◦ using Kriging with external drift (KED),
types dominating over high-, mid-, and low-elevation areas considering elevation as a predictor (Goovaerts, 2000). For
respectively. Permanent snow and ice cover a negligible area the precipitation, contrasting reviews are reported in the lit-
in the northwest of the basin, whereas a small patch of barren erature about the performance of the KED technique. For in-
land exists over the densely grassy mountains of the western stance, Masson et al. (2014) reported considerable improve-
Himalaya and Pir Panjal. The urban settlement covers a small ment in interpolation accuracy with KED compared to other
portion of the basin, concentrated in the Kashmir valley. linear regressions not accounting for any predictor in high
The climate of the UJB is influenced by dynamic local and mountainous regions. On the other hand, Berndt and Haber-
regional weather systems, and the topography of the high landt (2018) and Ly et al. (2011) argue that topographical
mountains causes a huge variability in the spatial and sea- impact was indispensable for only temperature reconstruc-
sonal distribution of precipitation (Dolk et al., 2020). Two tion at all temporal resolutions and station densities, but its
distinct precipitation patterns (i.e. western disturbances and influence was less clear for daily to monthly precipitation.
monsoon) exist in the basin. The western disturbances bring Furthermore, all spatial interpolation techniques can perform
precipitation in the form of snow during the winter sea- poorly in regions with insufficient high-elevation data, due
son. The monsoon pattern brings liquid rainfall during sum- to inaccurate estimation of local lapse rates (Ruelland and
mer seasons. The monsoon precipitation pattern dominates Sciences, 2020). Therefore, the ERA5 precipitation estimates
in the two lower sub-basins, i.e Poonch and Kanshi, and pro- (0.25◦ horizontal resolution) corrected for distribution map-
gressively loses strength northward towards the foothills of ping (DM) were used in the present study. ERA5 is a rela-
the western Himalaya, where the influence of western dis- tively new reanalysis launched by the European Centre for
turbances is predominant (Neelum and Kunhar sub-basins). Medium-Range Weather Forecasts (ECMWF) (Saha et al.,
The basin average annual precipitation and temperature is 2010). The data are developed by using an advanced 4D-
about 1150 mm yr−1 and 13.2 ◦ C respectively. Owing to the Var assimilation scheme and provide various atmospheric
steep rugged mountainous topography of the basin and con- variables at 139 pressure levels for the period 1979–present.
sequent short lag time, the flow level in the river and its tribu- The suitability of ERA5 to the UJB and surrounding region
taries rises abruptly during a rainfall event (Dar et al., 2019). was also reported by Liaqat et al. (2021) and Baudouin et
https://doi.org/10.5194/nhess-22-287-2022 Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022
290 R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features
Figure 1. Location of the UJB and spatial distribution of climatic stations.
al. (2020). The DM method adjusts the cumulative distri- tion of a variable x is expressed as
bution function (CDF) of modelled precipitation to match
with the observed precipitation CDF using a transfer func- " β #−1
α
tion (Sennikovs and Bethers, 2009), and it is commonly used F (x) = 1 + , (1)
to correct the systematic distributional biases (Cannon et al., x −γ
2015). The gamma distribution (Thom, 1958) with a shape
and a scale parameter was found to be suitable for the precip- where α, β, and γ are the shape, scale, and origin parame-
itation distribution in the study region (Azmat et al., 2018). ters respectively. In the second step, SPEI is calculated as the
The suitability of ERA5 precipitation and the bias correction standardized value of F (x) as follows:
method with respect to extreme precipitation analysis were
checked against observed station data, and a few results of C0 + C1 W + C2 W 2
SPEI = W − , (2)
the reliability check of DM-corrected ERA5 are provided in 1 + d1 W + d2 W 2 + d3 W 3
the Supplement (see Fig. S1).
where
p
4 Methods W= −2 ln (F (x)) for F (x) < 0.5 (3)
p
4.1 Wet and dry event identification W= −2 ln (1 − F (x)) for F (x) > 0.5. (4)
SPEI, the most widely used index, was adopted to detect and The parameters C0 , C1 , C2 , d1 , d2 , and d3 are SPEI con-
characterize wet and dry events of different severity levels stants (Vicente-Serrano et al., 2010). The log-logistic distri-
(moderate, severe, and extreme). The SPEI supports compar- bution for SPEI calculation was used and recommended by
isons over time and space, as proxies of wet and dry condi- many researchers (Ullah et al., 2021a; Akhtar et al., 2020;
tions from both the meteorological and agricultural perspec- Himayoun and Roshni, 2019; Vicente-Serrano et al., 2010).
tives. Although the SPEI was originally proposed for drought The detailed description of the SPEI calculation procedure
monitoring, it can also be used as a tool to detect flood risk. can be found in Vicente-Serrano et al. (2010). In this study,
The calculation procedure of SPEI involves two steps: fitting SPEI was calculated using the “SPEI” package in R environ-
a log-logistic distribution to the monthly climatic water bal- ment (Beguería et al., 2017). The severity levels of wet and
ance (P-PET) time series and then transforming the cumula- dry events based on SPEI values were classified according to
tive probability of the fitted distribution into a standard nor- Chen et al. (2020), and results are listed in Table 1. Positive
mal distribution (with mean zero and variance 1). According and negative values of SPEI represent the severity of wet and
to this distribution method, the probability distribution func- dry events respectively.
Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022 https://doi.org/10.5194/nhess-22-287-2022
R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features 291
Table 1. SPEI classification of dry and wet events (from Chen et al., 4.4 Wet–dry transition time
2020).
The total number of transitions and their average transition
SPEI value Description time (Tt ) in months for wet-to-dry and dry-to-wet events was
computed for each grid cell for the period 1981–2014, as de-
> 1.99 Extremely wet
1.99 to 1.50 Severely wet scribed by Luca et al. (2020). The calculation procedure of
1.49 to 1.00 Moderately wet wet-to-dry transition time (Tt ) involves four steps: (i) extrac-
0.99 to −0.99 Normal tion of wet and dry events and arranging them in an ascend-
−1.00 to −1.49 Moderately dry ing order of time (from the oldest to the most recent); (ii) in
−1.50 to −2.00 Severely dry case of consecutive dry and wet months, keep only the first
−2.00 < Extremely dry and the last month value respectively; (iii) calculate the dif-
ference in months between wet to dry events within the time
series; and (iv) take the average of the time interval. The same
4.2 Wet and dry event characteristics procedure was applied for calculating dry-to-wet transition
time (Tt ), with the only difference being in step (ii) in which
In this study, three characteristics (severity, duration, and in- the first and last months of wet and dry events were kept re-
tensity) of wet and dry events were calculated for each pixel. spectively, and in step (iii) in which the time interval was cal-
Following Spinoni et al. (2014), the duration (D) of a wet– culated between dry-to-wet events. The wet-to-dry and dry-
dry event is the length of time (months) that the index is to-wet transition times were calculated separately for each
consecutively above or below a truncation value; the severity level of severity (moderate, severe, extreme).
(S) refers to the cumulative value of the index from the first
month to the last month of the wet–dry event, and it repre- 4.5 Wet–dry rapid-transition events
sents the water surplus and deficit respectively; and the inten-
sity (I) of an event is the ratio of severity (S) to duration (D). The wet–dry rapid-transition event is defined as the consecu-
These characteristics were computed for each event and then tive occurrence of wet and dry months/events. For instance, a
further the total wet and dry event duration (TWD and TDD), dry (or wet) event occurring in the ith month abruptly altered
total wet and dry severity (TWS and TDS), total wet and dry to a wet (or dry) event in the i+1 month. In this study, the fre-
intensity (TWI and TDI), average wet and dry event duration quency of wet-to-dry (wet event followed by dry event) and
(AWD and ADD), average wet and dry severity (AWS and dry-to-wet (dry event followed by wet event) rapid-transition
ADS), average wet and dry intensity (AWI and ADI), max- events was calculated for each pixel to identify the geograph-
imum wet and dry event duration (MWD and MDD), maxi- ical hotspot for compound extreme events. Unlike the wet–
mum wet and dry severity (MWS and MDS), and maximum dry average transition time, which was calculated separately
wet and dry intensity (MWI and MDI) were calculated for a for each severity level, the wet–dry rapid-transition events
period of 34 years (1981–2014). were calculated considering all levels of severity together.
4.3 Wet–dry (WD) ratio
5 Results
The wet–dry (WD) ratio is defined as the natural logarithm
of the ratio of the total number of wet months (Nw ) to the to- 5.1 Change trends of the wet–dry events
tal number of dry months (Nd ) (Luca et al., 2020). The WD
The basin average SPEI time series at 1-month (SPEI-1),
ratio was calculated for different levels of severity (moder-
3-month (SPEI-3), 6-month (SPEI-6), and 12-month (SPEI-
ate, severe, and extreme) at each pixel for the studied period
12) timescales is presented in Fig. 2. It can be seen that the
(1981–2014) using Eq. (5):
study domain mostly experienced moderate-to-severe wet–
Nw
dry events, whereas the extreme wet–dry events (SPEI > 2
WD ratio = ln . (5) or SPEI < −2) rarely occurred during the study period. For
Nd
the SPEI-1, the wet (blue) and dry (red) events changed more
The WD ratio provides information about the susceptibility frequently than accumulated SPEI (at 3, 6, and 12 months),
of a given area to be more affected by wet or dry events. A and there was no extended dry or wet period. The reason
WD ratio greater than 0 implies the prevalence of wet events, might be that the precipitation and temperature of each new
whereas a WD ratio lower than 0 shows a dominance of dry month have a substantial impact on the accumulative val-
events. The natural logarithm was used to narrow the range ues of that period. By contrast, with the increase in SPEI
of WD ratio values and to separate the wet-dominated versus timescale (SPEI-1 to SPEI-12), a clear change/shift of basin
dry-dominated regions by sign. climate from wet to dry conditions can be seen (Fig. 2),
showing the stability in the frequency of incidence of wet–
dry events over the study domain. This could be explained
https://doi.org/10.5194/nhess-22-287-2022 Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022
292 R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features
as the slow and consistent response of SPEI towards changes In this study, the total, average, and maximum values of du-
in climatic variables, indicating strong and clear durations ration, severity, and intensity were computed for the study
of annual and multiple-year dry and wet conditions. This period (1981–2014). The maps of wet and dry duration are
means that at longer timescales of SPEI the number of oc- displayed in Fig. 4. Overall, the study area encountered rel-
currences of wet–dry events will decrease, but the duration atively more wet months than dry months during the whole
will increase. study period. The total wet duration (TWD) and the total dry
This study focuses on the short-timescale conditions to duration (TDD) vary from 66 to 80 and from 61 to 65 months
analyse frequent variations in climatic conditions and their respectively for most of the basin. The low-elevation parts
interplay; therefore, more detailed analysis was carried out in the south of the basin show the highest value of TWD
at the monthly timescale. Moreover, the floods and flash whereas the TDD is higher across the Himalaya divide line
droughts are not clearly associated with long-term SPEI, be- than in other parts of the basin. The Himalaya divide line is
cause the averaging effect of long-term accumulated precip- a line in the middle of the UJB at the Pir Panjal mountain-
itation and temperature surpasses the signal of extreme pre- ous range, separating the dominance of the two precipitation
cipitation and temperature over a short period. Flash drought patterns: westerlies in the north-facing slopes and monsoon
is a relatively new type of drought. Currently, there is not a in the south-facing slopes of the line (Archer and Fowler,
universally accepted definition or criteria for flash drought, 2008).
though there is general consensus on the principle of rapid The average wet and dry event durations (AWD and ADD)
onset or intensification characterized by moisture deficits and were found to be similar throughout the basin with a slight
abnormally high temperatures for a period lasting at least difference in the range of 1–2 weeks. However, their spatial
3 weeks (Lisonbee et al., 2021; Otkin et al., 2018; Hunt et al., patterns were found to be mostly complementary. Maximum
2009). This highlights the usefulness of SPEI at the monthly wet and dry event durations (MWD and MDD) exhibit high
scale in representing flood and flash drought events. It is values in two distinct parts of the basin. The MWD is about
noted that the terms “wet–dry events” or “wet–dry months” 6–7 months in the east of the basin, which is located in Kash-
present similar meaning for our study, as the analysis was mir, India, whereas it varies between about 4–5 months and
made at the monthly time step. A clearer picture of the 2–3 months in the northwest and southwest parts of the basin.
monthly evolution of wet–dry events of different severity lev- For the MDD, the northwest and central parts of the basin
els and their variability can be seen in Table 2. The SPEI-1 show higher values (4–5 months) than the remaining parts
values fluctuate remarkably from one month to another. For (2–3 months).
example, an extremely wet October in 1987 was followed by The spatial distribution of total, average, and maximum
a severely dry November, and a severely wet June occurred severity of wet–dry events is presented in Fig. 5. All wet–dry
at the tail of the longest drought spell in May 2001. Such severity maps show similar spatial patterns as wet–dry dura-
rapid transition from wet to dry and from dry to wet events tion maps. In terms of total wet severity (TWS) and total dry
was more prominent during the first half of the study period severity (TDS), the wet and dry hotspots are located in the
(before the year 1997). Another interesting observation con- south and middle (across Himalaya divide line) of the basin
cerns the strong change in the basin climatic features which respectively. Unlike the spatial patterns of TDD, the TDS is
can be noticed around the years 1997–1998. During the first relatively higher in the north of the basin above the Himalaya
half of the study period (1981–1997), the dominancy of wet divide line. This shows more intense dry events in this part
events of different categories prevails whereas the basin con- of the basin. The underlying reason for higher TDS could
ditions lean towards dryer conditions during the second half be the higher warming rates in western Himalaya, hosted in
of the period (1998–2014). the north of the basin. The average severity of wet and dry
Annual variations in the number of months affected by events is categorized from moderate to severe levels. The av-
dry–wet events (SPEI ≤ −1 and SPEI ≥ 1) is displayed in erage wet severity (AWS) exhibits random spatial patterns,
Fig. 3. Usually, every year encountered at least one dry and whereas the average dry severity (ADS) is relatively higher
wet month of any severity level. Approximately 35 % of in the north of the basin. Observed spatial patterns of maxi-
the total number of months experienced anomalous dry or mum wet severity (MWS) and maximum dry severity (MDS)
wet conditions. The proportion of wet months (18.1 %) was were similar to those of MWD and MDD. The eastern part of
slightly higher than that of dry ones (16.9 %). the basin experienced wet events of higher severity than the
western one, whereas the most severe dry events affected the
5.2 Wet–dry event analysis northwest and central parts of the basin.
Figures 6 illustrates the spatial distribution of intensities of
The wet–dry event characteristics (duration, severity, and in- wet–dry events, calculated as the ratio of severity to duration.
tensity) were computed for each pixel to analyse their spa- The total wet intensity (TWI) and total dry intensity (TDI)
tial distribution. Pixel-based analysis shows the location of vary from moderate to severe with a noted range of 1.44 to
the most vulnerable parts of the basin, providing the basis 1.55 and −1.36 to −1.52 for wet and dry events respectively.
for future decisions on adaptation and mitigation measures. Irrespective of TWD and TWS, which is highest in the south
Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022 https://doi.org/10.5194/nhess-22-287-2022
R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features 293 Figure 2. Temporal variations in SPEI at 1-, 3-, 6-, and 12-month timescales over UJB for the period 1981–2014. Figure 3. Annual variations in the number of months affected by wet–dry conditions during 1984–2014. The brown and blue colours present dry and wet months respectively. Different shades of the colours define the different severity levels (EW – extreme wet; ED – extreme dry; SW – severe wet; SD – severe dry; MW – moderate wet; MD – moderate dry). https://doi.org/10.5194/nhess-22-287-2022 Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022
294 R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features
Table 2. Temporal variations in monthly SPEI over UJB from 1981–2014. The blank cells show normal months, and the different severity
levels are presented as EW – extremely wet; ED – extremely dry; SW – severely wet; SD –severely dry; MW – moderately wet; and MD –
moderately dry). The line between 1997 and 1998 indicates the strong change in the basin climatic features.
Year/months 1 2 3 4 5 6 7 8 9 10 11 12
1981 MW
1982 MD SD MW SW MW
1983 SW MW MW
1984 MW MD
1985 SD MD MW MD SW
1986 MD SW MW SW MW
1987 MD MW EW SD MD EW SD
1988 MW MD MD SW MD
1989 MW MW
1990 MD EW
1991 SW MW SD MW
1992 SW SW MD SW
1993 MW SD SW ED MW MD
1994 EW MW SW MD SW
1995 SW EW MW MD
1996 MW SW EW MW SD SW
1997 MD MW MW SW MW
1998 MW SD SD
1999 MW MD SD MW SD
2000 MD MD SD SD MD
2001 SD SD MD SD SW MW
2002 MD SD MD
2003 MD SW MD
2004 MD SD MD MD MW
2005 EW SD MD
2006 SW MD SD SW MW MW
2007 SD ED SW MD SD MD
2008 SW SD MW MW
2009 MD SD
2010 MD SW MD MW SW MW MD MD
2011 SW MD MW
2012 SD SD SW
2013 MD MD EW
2014 MW EW MD
of the basin, TWI is more intense in the middle and north- 5.3 Wet–dry ratio
east of the basin. The TDI is found to be more intense over
western Himalaya, north of the basin. The average wet in-
tensity (AWI) and average dry intensity (ADI) vary within The WD ratio features the dominance of wet or dry events
the moderate class of hazard. However, their spatial patterns for the period of 34 years (1981–2014). The WD ratio for the
are much different from average duration (AWD and ADD) three severity levels (moderate, severe, and extreme) at pixel
and average severity (AWS and ADS) patterns. Regarding basis is presented in Fig. 7. The positive and negative values
maximum intensities, the spatial patterns of maximum wet of WD ratio depict the prevalence of wet and dry events re-
intensity (MWI) resemble the patterns of MWD and MWS spectively. As the figure shows, higher frequencies of mod-
well, whereas the maximum dry intensity (MDI) exhibits erate dry events with respect to moderate wet events were
much different spatial patterns from MDD and MDS. The found throughout the basin except for a few pixels in the
dry events are found to be more intense than wet events, but south. By contrast, severe to extreme wet events are more
only for a few pixels in the southwest of the basin. On the frequent for most parts of the basin. The highest positive val-
other hand, wet events with higher intensities are found to be ues of WD ratio for extreme level of hazard were found in the
more widespread than dry events. southwest of the basin, which shows the higher susceptibility
of the area towards extreme wet events. Moreover, the anal-
Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022 https://doi.org/10.5194/nhess-22-287-2022R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features 295
Figure 4. Spatial distribution of total wet duration (TWD), total dry
Figure 6. Spatial distribution of total wet intensity (TWI), total dry
duration (TDD), average wet duration (AWD), average dry duration
intensity (TDI), average wet intensity (AWI), average dry intensity
(ADD), maximum wet duration (MWD), and maximum dry dura-
(ADI), maximum wet intensity (MWI), and maximum dry intensity
tion (MDD) for the period 1981–2014.
(MDI) for the period 1981–2014.
Figure 7. Spatial distribution of wet–dry (WD) ratio derived for
three levels of severity (moderate, severe, and extreme) during the
period 1981–2014. Blue (WD ratio > 0) means that the area expe-
rienced more wet than dry events. Brown (WD ratio < 0) indicates
the opposite.
2014 are presented in Figs. 8 and 9. As expected, the num-
Figure 5. Spatial distribution of total wet severity (TWS), total dry ber of transitions for wet-to-dry and dry-to-wet events was
severity (TDS), average wet severity (AWS), average dry severity the highest for the moderate level of events, followed by se-
(ADS), maximum wet severity (MWS), and maximum dry severity
vere and extreme levels of events. Consequently the average
(MDS) for the period 1981–2014.
transition time from wet-to-dry and dry-to-wet events was
found to be the highest for the extreme level of event fol-
ysis of wet–dry event characteristics also revealed the preva- lowed by severe and moderate levels of events. The num-
lence of wet events with higher duration and severity over ber of transitions for moderate, severe, and extreme levels of
monsoon-dominated regions. events varies from 15 to 26, from 6 to 16, and from 1 to 5 re-
spectively. Overall, the number of transitions for dry-to-wet
5.4 Wet–dry transition time events is larger than the wet-to-dry events for severe and ex-
treme levels of events, whereas the opposite was found for
The number of transitions and their average transition time the moderate level of events. The transition time for mod-
for wet-to-dry and dry-to-wet events for the period 1981– erate, severe, and extreme levels of events varies from 1.8
https://doi.org/10.5194/nhess-22-287-2022 Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022296 R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features
Figure 9. Average transition time (Tt ) intervals in months for wet-
Figure 8. Number of transitions from wet-to-dry (left) and dry-to- to-dry (left) and dry-to-wet (right) events for three levels of severity
wet (right) events for three levels of severity (moderate, severe, ex- (moderate, severe, extreme) for the period 1981–2014.
treme) for the period 1981–2014.
to 6.5, from 1.8 to 16.75, and from 3.5 to 187.0 months re-
spectively. Overall, 53.57 % and 17.86 % of pixels in the UJB
showed longer transition time from wet to dry than from dry
to wet for moderate and extreme levels, whereas the opposite
was seen for severe events.
5.5 Wet–dry rapid-transition events
The wet–dry rapid transition is the consecutive occurrence
of wet and dry months of any severity level. The fre-
quency of wet-to-dry (wet month followed by dry month)
and dry-to-wet (dry month followed by wet month) rapid-
transition events was computed for each grid cell and is Figure 10. Frequency of occurrence of abrupt events, wet to dry
shown in Fig. 10. The frequency of wet–dry transition events (left) and dry to wet (right), during the period 1981–2014.
varies/ranges from 5 to 20 events during the 34 years of the
study period. About 50 % of pixels in the UJB encountered a
higher number of wet events terminated at dry months. The 6 Discussion and conclusion
spatial distribution of frequency of wet–dry rapid-transition
events revealed that the wet-to-dry events are less frequent This study attempts to investigate the spatiotemporal varia-
over the westerlies-dominated region of the basin, whereas tions in wet–dry events collectively, their characteristics (du-
the southwestern part of the basin was more affected by ration, severity, intensity), and transition from wet-to-dry and
the abrupt wet-to-dry events. By contrast, abrupt dry-to-wet dry-to-wet events during the period 1981–2014 in the Up-
events are found to be more frequent over pixels surrounding per Jhelum Basin (UJB) in South Asia. The SPEI, which
the Himalaya divide line, whereas the remaining part of the incorporates both precipitation and potential evapotranspira-
basin depicts less incidence of dry-to-wet events. tion, was used to extract and analyse the wet–dry events. The
Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022 https://doi.org/10.5194/nhess-22-287-2022R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features 297 whole analysis was carried out at the monthly timescale, but significant decrease in precipitation extremes over Southeast the temporal evolution of the basin-averaged index was also Asia, Indonesia, Australia, and the northernmost region of simulated at multiple timescales (1, 3, 6, and 12 months). The South America during El Niño phases, whereas in the south- reason for selecting the monthly timescale for this study is ern tier of the United States and the region from Argentina to that it is expected to provide the best performance in detect- southern Brazil heavy precipitation increased during El Niño ing floods and flash droughts, as longer time steps are more phases, and vice versa during La Niña phases. The strength appropriate for long-term droughts only and not for floods. of such connections for Pakistan was also demonstrated in The results of temporal variations in SPEI showed that the several studies. El Niño suppresses monsoon rainfall activity study domain mostly encountered moderate to severe wet– over Pakistan, while La Nina has a negative impact on win- dry events, whereas the extreme wet–dry events rarely oc- ter precipitation over Pakistan (Farooqi et al., 2005; Khan, curred during the study period. The results of basin-average 2004). Ullah et al. (2021a) found significant impacts of three SPEI at multiple timescales revealed that the response of large-scale climate indices, i.e sea surface temperature (SST) SPEI to the deviations in climatic features varies with the and multivariate El Niño–Southern Oscillation (ENSO4.0), accumulation time. Therefore, shorter timescales are more on seasonal droughts across Pakistan. appropriate for detecting frequent seasonal and inter-annual The results of wet–dry event characteristics (duration, variations, whereas longer timescales provide useful infor- severity, intensity) at pixel basis outline the greater suscep- mation regarding the signature of the events over the region tibility of the westerlies-dominated region to dry events with (Ayugi et al., 2020; Du et al., 2013). Furthermore, the SPEI higher duration, severity, and intensity. The dryer condi- time-series plots capture the observed extreme floods and tions in this region could be explained with the increasing drought events that occurred in the basin during the study rates of global warming over the mountainous region of the period well: for instance, the longest drought event occurred basin, also reported by many researchers (Rashid et al., 2020; from the late 1990s to early 2000s, as evident in Fig. 2 and Shafiq et al., 2020; Zaz et al., 2019). Studies by Negi et al. Table 2. The drought started in 1998 and was considered to (2018) and Dimri and Dash (2012) also confirm that most of be the worst in the history of Pakistan. The drought spell the western Himalayan region recorded a significant warm- in 2001–2002 resulted in water shortage of up to 51 % of ing trend from 1975 onwards in particular. This is also sup- normal supplies (Ahmad et al., 2004). Likewise, the notable ported by the tree-ring chronologies of the region, which in- flooding events, usually flash floods ranging from moder- dicate a rapid growth of the tree rings in recent decades, es- ate to severe, occurred in the years 1988, 1992, 1994, 1997, pecially at higher altitudes (Borgaonkar et al., 2009). The im- 2007, and 2014 (Bhat et al., 2019) and were well captured by pact of global warming on short-term dry events (soil mois- the SPEI, confirming its valuable contribution to this type of ture drought) is not straightforward as rising temperature did analysis. not necessarily cause increase in actual ET, especially in arid An interesting clue to the changing climate is the strong and semiarid regions (Trenberth et al., 2014; Sheffield et al., change that occurred in the basin at the end of 1997 (Table 2). 2012). In fact the rate and amount of ET results from a com- Before this change (1981–1997), wet events of different plex interaction of temperature, radiation balance, precipita- severity levels predominated in the basin, whereas dryer con- tion rates and vegetation physiological control, rather than ditions prevailed after 1997. However, it still needs to be in- being exclusively limited by one of these factors. For flash vestigated whether dryer conditions are expected to continue drought, the rapid soil moisture decline should be a result in the future or whether a large multi-decadal variation is tak- of the intensification of ET driven by higher temperature, ing place. This strong change in the basin climate coincides which is very common in humid and semi-humid regions, with the strongest El Niño–Southern Oscillation (ENSO) where soil moisture can sustain higher ET amounts up to a event in the winter season of 1997–1998, where the Oceanic few weeks (Yuan et al., 2019). Further decrease in winter Niño Index (ONI) peaked at 2.3 and influenced the climate and spring precipitation leads to water deficit conditions in conditions all over the world (MRCC, 2021). The 1998–2002 this part of the basin. The worst drought event period (2000– drought in southwestern Asia, accompanied by the most se- 2001), partially induced by a stronger ENSO in winter, was vere drought conditions in the last 50 years, was also a re- also due to the low winter and spring precipitation, as shown sult of this strong ENSO event (Ain et al., 2020; Ahmed et in Table 2. During 2000–2001, winter and spring seasons al., 2018). ENSO is the primary mode of inter-annual vari- were moderately to severely dry, whereas the monsoon and ability, having great influence on global weather and climate autumn seasons observed normal months. By contrast, the via atmospheric circulations (Ullah et al., 2021a). Many re- higher duration and severity of wet events were detected in searchers reported the close association between variations in the monsoon-dominated region, implying that floods mainly atmospheric circulation patterns and climatic variables, ex- occurred during monsoon season with heavy rainfall along treme weather phenomena like drought and flood (Luca et with snowmelt. However, the eastern part of the basin was al., 2020; Omidvar et al., 2016; Sun et al., 2015). Kenyon the hotspot of more intense wet events. The above discussion and Hegerl (2010) examined the response patterns of hydro- is also supported by the historic database of observed flood climate extremes to ENSO over global land areas and stated a https://doi.org/10.5194/nhess-22-287-2022 Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022
298 R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features
events, as most of these events occurred during monsoon sea- source R packages. The potential evapotranspiration (PET) and
son. standardized precipitation evapotranspiration index (SPEI) were
The results of the WD ratio showed the prevalence of se- calculated with the R package “SPEI” (version 1.7). The interpo-
vere to extreme wet events for most of the basin, while the lation of PET values was done using the R package “hydroTSM”
dry events of moderate severity level were more frequent in (Version 0.6-0). Data and code written to generate the figures are
available from the corresponding author upon request.
the study domain. The southwestern part of the basin, lo-
cated in the monsoon-dominated region, was found to be
the hotspot for the extreme wet events. Moreover, the analy-
Data availability. The ERA5 precipitation data can be accessed
sis of wet–dry event characteristics also revealed the preva-
online (DOI: https://doi.org/10.24381/cds.adbb2d47, Hersbach et
lence of wet events with higher duration and severity over al., 2018). For the Pakistani region, the observed precipitation and
the same monsoon-dominated region. The spatial patterns of temperature data are available from the Pakistan Meteorological
average transition time from one extreme type to the other Department (PMD) and Water and Power Development Author-
type was found to be heterogeneous and different for the ity (WAPDA) upon request. For the Indian region, Indian Meteo-
three severity levels. Overall, a greater number of pixels took rological Department (IMD) daily gridded precipitation and tem-
a shorter time to switch from dry to wet events than from perature datasets are freely available (https://cdsp.imdpune.gov.in/
wet to dry events. Apart from the average transition period, home_gridded_data.php, Pai et al., 2014).
the study domain also experienced rapid transition of wet–
dry events. In general, the surrounding region of the Hi-
malaya divide line and the monsoon-dominated part of the Supplement. The supplement related to this article is available on-
basin were found to be the hotspots of rapid wet–dry tran- line at: https://doi.org/10.5194/nhess-22-287-2022-supplement.
sition. The rapid wet–dry swings could be explained in the
context of global warming. In a warmer climate, increased
Author contributions. This paper was conceptualized by RA and
evapotranspiration rates in response to increased temperature
GG. RA performed the data analysis and visualization. The original
could elevate the drought risk and frequency. At the same draft was written by RA and revised by GG.
time, the prospect of localized heavy precipitation causing
floods is expected to increase in response to increased atmo-
spheric moisture content due to increased evapotranspiration Competing interests. The contact author has declared that neither
rates (He and Sheffield, 2020; Krishnan et al., 2020). Further they nor their co-authors have any competing interests.
warming-induced changes in global climate variability, such
as El Niño and La Niña, can cause more inter-annual variabil-
ity or persistence in global weather and climate, significantly Disclaimer. Publisher’s note: Copernicus Publications remains
affecting regional precipitation and temperature distribution neutral with regard to jurisdictional claims in published maps and
in space and time (Ullah et al., 2021b). Further compelling institutional affiliations.
scientific evidence of human interventions, such as boosted
human water intake and land use changes, exacerbates the
extreme flood and drought risk hazard. Special issue statement. This article is part of the special issue “Re-
To conclude, knowledge of wet–dry event characteristics cent advances in drought and water scarcity monitoring, modelling,
and their rapid transition provides meaningful insight into the and forecasting (EGU2019, session HS4.1.1/NH1.31)”. It is a re-
geographical hotspots of compound extreme events, which sult of the European Geosciences Union General Assembly 2019,
Vienna, Austria, 7–12 April 2019.
could be of practical value to inform a group of stakeholders
(researchers, local authorities, policy makers, relief agencies,
non-governmental organizations (NGOs), and (re)insurance
Acknowledgements. The authors are grateful to the Pakistan Mete-
companies) on the potential risk. In general, results con- orological Department (PMD) and Water and Power Development
tribute to hydrological predictability and risk assessment and Authority (WAPDA) for sharing the data.
therefore effectively support disaster preparedness and risk
management, ensuring the regional water, food, and socio-
economic security and stability against the background of a Financial support. The authors received funding from the Cooper-
changing environment. Future work should explore to what ation Agreement PFK PhD programme 2019–2022 “Partnership for
extent future wet–dry event frequency will respond to anthro- Knowledge-Platform 2: Health and WASH (WAter Sanitation and
pogenic forcing, internal atmospheric processes, and human good Hygiene)” of the AICS-Italian Agency for Development Co-
interventions. operation to attend higher education programmes in Italy in favour
of non-Italian citizens.
Code availability. All calculations and plots were produced using
ArcMap (version 10.8) and R (version3.3.2) by making use of open-
Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022 https://doi.org/10.5194/nhess-22-287-2022R. Ansari and G. Grossi: Spatio-temporal evolution of wet–dry event features 299
Review statement. This paper was edited by Brunella Bonaccorso Azmat, M., Qamar, M. U., Huggel, C., and Hussain, E.:
and reviewed by two anonymous referees. Future climate and cryosphere impacts on the hydrol-
ogy of a scarcely gauged catchment on the Jhelum river
basin, Northern Pakistan, Sci. Total Environ., 639, 961–976,
https://doi.org/10.1016/j.scitotenv.2018.05.206, 2018.
Baudouin, J.-P., Herzog, M., and Petrie, C. A.: Cross-validating
References precipitation datasets in the Indus River basin, Hydrol. Earth
Syst. Sci., 24, 427–450, https://doi.org/10.5194/hess-24-427-
Ahmad, S., Hussain, Z., Qureshi, A. S., Majeed, R., and Saleem, 2020, 2020.
M.: Drought mitigation in Pakistan: current status and options Beguería, S., Vicente-Serrano, S. M., Reig, F., and Latorre, B.: Stan-
for future strategies, vol. 85, IWMI, ISBN 92-9090-580-8, 2004. dardized precipitation evapotranspiration index (SPEI) revis-
Ahmed, K., Shahid, S., and Nawaz, N.: Impacts of cli- ited: parameter fitting, evapotranspiration models, tools, datasets
mate variability and change on seasonal drought char- and drought monitoring, Int. J. Climatol., 34, 3001–3023,
acteristics of Pakistan, Atmos. Res., 214, 364–374, https://doi.org/10.1002/joc.3887, 2014.
https://doi.org/10.1016/j.atmosres.2018.08.020, 2018. Beguería, S., Vicente-Serrano, S. M., and Beguería, M. S.: Pack-
Ain, N., Latif, M., Ullah, K., Adnan, S., Ahmed, R., Umar, age “SPEI”, Calculation of the Standardised Precipitation-
M., and Azam, M.: Investigation of seasonal droughts and Evapotranspiration Index, CRAN [Package], 2017.
related large-scale atmospheric dynamics over the Potwar Berndt, C. and Haberlandt, U.: Spatial interpolation of climate vari-
Plateau of Pakistan, Theor. Appl. Climatol., 140, 69–89, ables in Northern Germany – Influence of temporal resolution
https://doi.org/10.1007/s00704-019-03064-8, 2020. and network density, Journal of Hydrology: Regional Studies,
Akhtar, T., Mushtaq, H., and Hashmi, M. Z.-R.: Drought monitor- 15, 184–202, https://doi.org/10.1016/j.ejrh.2018.02.002, 2018.
ing and prediction in climate vulnerable Pakistan: Integrating hy- Bhat, M. S., Alam, A., Ahmad, B., Kotlia, B. S., Farooq, H.,
drologic and meteorologic perspectives, Hydrol. Earth Syst. Sci. Taloor, A. K., and Ahmad, S.: Flood frequency analysis of
Discuss. [preprint], https://doi.org/10.5194/hess-2020-297, in re- river Jhelum in Kashmir basin, Quatern. Int., 507, 288–294,
view, 2020. https://doi.org/10.1016/j.quaint.2018.09.039, 2019.
Ali, Z., Hussain, I., and Faisal, M.: Annual Characterization of Re- Borgaonkar, H., Ram, S., and Sikder, A.: Assessment of tree-
gional Hydrological Drought using Auxiliary Information under ring analysis of high-elevation Cedrus deodara D. Don
Global Warming Scenario, Nat. Hazards Earth Syst. Sci. Discuss. from Western Himalaya (India) in relation to climate
[preprint], https://doi.org/10.5194/nhess-2018-373, 2019. and glacier fluctuations, Dendrochronologia, 27, 59–69,
Arias, P. A., Bellouin, N., Coppola, E., Jones, R. G., Krinner, G., https://doi.org/10.1016/j.dendro.2008.09.002, 2009.
Marotzke, J., Naik, V., Palmer, M. D., Plattner, G.-K., Rogelj, J., Cannon, A. J., Sobie, S. R., and Murdock, T. Q.: Bias correction
Rojas, M., Sillmann, J., Storelvmo, T., Thorne, P. W., Trewin, B., of GCM precipitation by quantile mapping: How well do meth-
Achuta Rao, K., Adhikary, B., Allan, R. P., Armour, K., Bala, ods preserve changes in quantiles and extremes?, J. Climate, 28,
G., Barimalala, R., Berger, S., Canadell, J. G., Cassou, C., Cher- 6938–6959, https://doi.org/10.1175/JCLI-D-14-00754.1, 2015.
chi, A., Collins, W., Collins, W. D., Connors, S. L., Corti, S., Chen, H., Wang, S., Zhu, J., and Zhang, B.: Projected
Cruz, F., Dentener, F. J., Dereczynski, C., Di Luca, A., Niang, A. changes in abrupt shifts between dry and wet extremes
D., Doblas-Reyes, F. J., Dosio, A., Douville, H., Engelbrecht, F., over China through an ensemble of regional climate model
Eyring, V., Fischer, E., Forster, P., Fox-Kemper, B., Fuglestvedt, simulations, J. Geophys. Res.-Atmos., 125, e2020JD033894,
J. S., Fyfe, J. C., Gillett, N. P., Goldfarb, L., Gorodetskaya, I., https://doi.org/10.1029/2020JD033894, 2020.
Gutierrez, J. M., Hamdi, R., Hawkins, E., Hewitt, H. T., Hope, P., CSIRO, State of the Climate 2018, Commonwealth Scientific
Islam, A. S., Jones, C., Kaufman, D. S., Kopp, R. E., Kosaka, Y., and Industrial Research Organisation, Australia, ISBN 978-1-
Kossin, J., Krakovska, S., Lee, J.-Y., Li, J., Mauritsen, T., May- 925315-97-4, 2018.
cock, T. K., Meinshausen, M., Min, S.-K., Monteiro, P. M. S., Dar, R. A., Mir, S. A., and Romshoo, S. A.: Influence of geomorphic
Ngo-Duc, T., Otto, F., Pinto, I., Pirani, A., Raghavan, K., Ranas- and anthropogenic activities on channel morphology of River
inghe, R., Ruane, A. C., Ruiz, L., Sallée, J.-B., Samset, B. H., Jhelum in Kashmir Valley, NW Himalayas, Quatern. Int., 507,
Sathyendranath, S., Seneviratne, S. I., Sörensson, A. A., Szopa, 333–341, https://doi.org/10.1016/j.quaint.2018.12.014, 2019.
S., Takayabu, I., Tréguier, A.-M., van den Hurk, B., Vautard, R., De Luca, P., Messori, G., Wilby, R. L., Mazzoleni, M., and
von Schuckmann, K., Zaehle, S., Zhang, X., and Zickfeld, K.: Di Baldassarre, G.: Concurrent wet and dry hydrological ex-
Technical Summary, in: Climate Change 2021: The Physical Sci- tremes at the global scale, Earth Syst. Dynam., 11, 251–266,
ence Basis, Contribution of Working Group I to the Sixth Assess- https://doi.org/10.5194/esd-11-251-2020, 2020.
ment Report of the Intergovernmental Panel on Climate Change, Dimri, A. and Dash, S.: Wintertime climatic trends in the
edited by: Masson-Delmotte, V., Zhai, P., Pirani, A., Connors, S. western Himalayas, Climatic Change, 111, 775–800,
L., Péan, C., Berger, S., Caud, N., Chen, Y., Goldfarb, L., Gomis, https://doi.org/10.1007/s10584-011-0201-y, 2012.
M. I., Huang, M., Leitzell, K., Lonnoy, E., Matthews, J. B. R., Dolk, M., Penton, D. J., and Ahmad, M. D.: Amplification of
Maycock, T. K., Waterfield, T., Yelekçi, O., Yu, R., and Zhou, hydrological model uncertainties in projected climate simula-
B., Cambridge University Press, in press, 2022. tions of the Upper Indus Basin: Does it matter where the
Ayugi, B., Tan, G., Niu, R., Dong, Z., Ojara, M., Mumo, L., water is coming from?, Hydrol. Process., 34, 2200–2218,
Babaousmail, H., and Ongoma, V.: Evaluation of meteorologi- https://doi.org/10.1002/hyp.13718, 2020.
cal drought and flood scenarios over Kenya, East Africa, Atmo-
sphere, 11, 307, https://doi.org/10.3390/atmos11030307, 2020.
https://doi.org/10.5194/nhess-22-287-2022 Nat. Hazards Earth Syst. Sci., 22, 287–302, 2022You can also read
